pggan-pytorch
pggan-pytorch copied to clipboard
Is there some way to download the images for Celeba-HQ
Or if you can make them available to me I can host them somewhere for people to download. I'd really like the dataset but don't particularly want to go through the process to generate it myself.
The authors released the dataset, and here is the link https://drive.google.com/open?id=0B4qLcYyJmiz0TXY1NG02bzZVRGs
please refer to this repo . https://github.com/tkarras/progressive_growing_of_gans
@nashory seems it is only the pretrained model, not the source images
@petergerten aren't they stored in the .dat files in the datasets directory? https://drive.google.com/drive/folders/1iGK3jDe_1-V0Lyy5468WNpb7y74u-iEp
The .dat files from google drive only contain the difference between the celebA and the celebA-HQ images. The process to create the images is actually quite complicated.
I just made my scripts available at
https://github.com/nperraud/download-celebA-HQ
They are far from perfect but I hope it helps.
Hi @nperraud! Thanks for sharing the scripts. I am trying to download the deltas from Google Drive but it gives me an error that the "download quota exceeded for this file" for all files of course. I guess there is a limit to the number of downloads per day for Google Drive. Do you have any idea if I can find the files stored somewhere else?
I generated the files and stored it as convenient zip files. You can download from Google Drive or generate using pre-built docker image. Visit suvojit-0x55aa/celebA-HQ-dataset-download
@suvojit-0x55aa there should be 200k images in Celeba HQ,but your sharing are 30k of them.
@shartoo As quoted in the Progressive Growing of GANs paper
In this section we describe the process we used to create the high-quality version of the CELEBA dataset, consisting of 30000 images in 1024 × 1024 resolution.
The Celeba HQ dataset contains only 30k images.
sorry for my careless check,you're right.
Another question,how could i use the list_attr_celaba.txt namely attribution label for celaba-hq.Does the 30k images of celeba-hq are first 30k of 200k images in celaba dataset?
Got it ,there is a image list file https://raw.githubusercontent.com/nperraud/download-celebA-HQ/master/image_list.txt which releated celaba and celaba-hq files.
@shartoo As quoted in the Progressive Growing of GANs paper
In this section we describe the process we used to create the high-quality version of the CELEBA dataset, consisting of 30000 images in 1024 × 1024 resolution.
The Celeba HQ dataset contains only 30k images.
Is It possible to share labels with the dataset? Thanks!
@Cuilie you can use this file.
@suvojit-0x55aa On your github page you said "The size of the final dataset is 89G". But in Google Drive, you have far less than 89Gb. How come?
@aciobanusebi as far as I remember that's the memory needed for raw dataset, the deltas, and the generated dataset.